{ "run_name": "R4_softce_nat_s0", "cfg": { "run_name": "R4_softce_nat_s0", "tokenizer": "vq", "patch": 32, "ctx_span": 4096, "fine_span": 1024, "enc_hidden": 64, "fsq_levels": [ 8, 6, 5, 5, 5 ], "vq_codes": 1024, "vq_dim": 64, "vq_beta": 0.25, "vq_decay": 0.99, "vq_entropy_w": 0.0, "vq_warm_steps": 2000, "vq_warm_decay": 0.9, "aux_code_ce": 0.2, "aux_code_offsets": [ 1 ], "aux_side_codes": false, "ce_smooth_eps": 0.1, "ce_smooth_k": 8, "aux_horizon_ce": 0.0, "d_model": 256, "n_layers": 4, "n_heads": 4, "train_kmin": 4, "train_kmax": 32, "cpm_cmax": 16, "cpm_pmax": 0.4, "cpm_span_max": 4, "steps": 30000, "batch": 64, "n_variates": 8, "lr": 0.001, "normuon_lr": 0.0008, "weight_decay": 0.0001, "grad_clip": 1.0, "warmup_frac": 0.03, "decay_frac": 0.1, "seed": 0, "data_seed": 0, "pool": "short", "eval_every": 500, "ckpt_every": 2500, "gift_ckpts": [ 7500, 15000, 22500, 30000 ] }, "val_log": { "step": [ 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, 10500, 11000, 11500, 12000, 12500, 13000, 13500, 14000, 14500, 15000, 15500, 16000, 16500, 17000, 17500, 18000, 18500, 19000, 19500, 20000, 20500, 21000, 21500, 22000, 22500, 23000, 23500, 24000, 24500, 25000, 25500, 26000, 26500, 27000, 27500, 28000, 28500, 29000, 29500, 30000 ], "val_crps": [ 0.8672901391983032, 0.9543266296386719, 0.9063344597816467, 1.010547161102295, 1.132070541381836, 0.919248640537262, 0.949482798576355, 0.8466875553131104, 1.1995621919631958, 0.91310054063797, 0.9678660035133362, 0.9958806037902832, 1.1102042198181152, 1.028991937637329, 0.8864909410476685, 1.0250556468963623, 1.1381415128707886, 1.0111591815948486, 0.9892610311508179, 1.0417277812957764, 0.8260747194290161, 0.9019104242324829, 0.8371614813804626, 0.9387744069099426, 0.9849492311477661, 0.8855383992195129, 0.9914988279342651, 0.9312825798988342, 1.0766913890838623, 0.884867787361145, 0.8108775615692139, 0.9834892749786377, 0.8895864486694336, 0.8799145817756653, 0.8217504024505615, 0.8388699889183044, 0.8132870197296143, 0.837430477142334, 0.8259345889091492, 0.8411772847175598, 0.7920902967453003, 0.899206817150116, 0.786103367805481, 0.8689879179000854, 0.8376018404960632, 0.8194923400878906, 0.8323427438735962, 0.8978451490402222, 0.8828029036521912, 0.910976231098175, 0.8309593200683594, 0.7401626706123352, 0.7400513887405396, 0.848382830619812, 0.8314951658248901, 0.7583629488945007, 0.9393184185028076, 0.7455391883850098, 0.7827884554862976, 0.765039324760437 ], "train_loss": [ 0.9604882258057234, 1.184627042033105, 1.2307118536590431, 1.298633979143299, 1.2711060607480096, 1.2986351054721506, 1.2652323591568322, 1.2466889629666478, 1.221811203037111, 1.2442857059692005, 1.2818746689783587, 1.2619748978864673, 1.2642940206737179, 1.2693755435805258, 1.1991855002078162, 1.2383370550126152, 1.233168522472853, 1.22721501932427, 1.213815648861966, 1.1972596122003443, 1.224107148493037, 1.2567336804814848, 1.1829379268755227, 1.2023012271003934, 1.2198481781301878, 1.2162967905452382, 1.1634902025369611, 1.202069006152197, 1.1717159303746398, 1.1574958028837354, 1.1428914341738492, 1.1749087183631945, 1.1799220188928925, 1.2062743716148177, 1.1706217073332668, 1.1959618292051606, 1.1454995063576208, 1.1439626029816452, 1.1262470030420266, 1.16410320163329, 1.1493091716000783, 1.1472210375481535, 1.1514362618817984, 1.159293913321349, 1.1437845871022971, 1.1639744104237582, 1.1366997163370811, 1.1648353799457543, 1.15462348002567, 1.172037148777154, 1.1562121463293746, 1.1411964291389873, 1.1555817428594217, 1.165637157662266, 1.15174152785994, 1.1101811476105479, 1.1105637267078636, 1.0948799233947506, 1.1116025525041666, 1.1357420209550548 ], "perplexity": [ 317.3086242675781, 381.8880310058594, 406.2688903808594, 266.072265625, 216.48526000976562, 115.4115219116211, 73.9161148071289, 139.71795654296875, 151.41455078125, 95.16780090332031, 63.69728088378906, 90.45946502685547, 71.08965301513672, 77.7764663696289, 172.1228790283203, 89.02313995361328, 109.26634979248047, 125.96153259277344, 86.3282699584961, 176.51959228515625, 143.86712646484375, 91.70943450927734, 167.63995361328125, 225.47581481933594, 115.31745147705078, 160.1117401123047, 133.34579467773438, 105.9024658203125, 149.9974365234375, 328.2115173339844, 351.7203369140625, 269.8708190917969, 217.7549591064453, 195.59548950195312, 152.72061157226562, 199.35513305664062, 169.0142364501953, 353.00518798828125, 270.88311767578125, 314.8919982910156, 365.21612548828125, 367.2245178222656, 433.7919616699219, 329.8296813964844, 270.558349609375, 425.4227600097656, 288.9580993652344, 294.4505310058594, 453.3986511230469, 366.81219482421875, 381.4461669921875, 379.4607849121094, 212.17196655273438, 189.31044006347656, 391.77581787109375, 368.7718811035156, 293.7601623535156, 373.07257080078125, 494.1739501953125, 553.6787109375 ], "code_ce": [ 3.0652949810028076, 4.484424114227295, 4.236876964569092, 4.532713413238525, 4.463877201080322, 4.253510475158691, 4.0902533531188965, 3.904784917831421, 4.240498065948486, 4.348794460296631, 4.15089225769043, 4.496209144592285, 4.228549957275391, 4.006947040557861, 4.493416786193848, 3.642125368118286, 4.072838306427002, 4.133636474609375, 4.5023112297058105, 4.120343208312988, 4.239618301391602, 4.4892191886901855, 4.094030857086182, 3.9324824810028076, 4.04104471206665, 4.360285758972168, 3.597569704055786, 4.037103652954102, 4.10603141784668, 3.9713797569274902, 4.0599260330200195, 3.7781097888946533, 3.9975008964538574, 3.9301629066467285, 4.012686729431152, 3.825331926345825, 3.8277981281280518, 3.9652748107910156, 3.642442226409912, 3.9585163593292236, 3.9427382946014404, 3.9965224266052246, 3.924818992614746, 3.98770809173584, 3.9484479427337646, 4.010968208312988, 3.7739217281341553, 3.988572835922241, 4.042597770690918, 4.086676120758057, 3.912381410598755, 3.963050603866577, 3.711103677749634, 3.8999063968658447, 4.022610664367676, 3.882422685623169, 3.7607550621032715, 3.7202415466308594, 3.8536312580108643, 3.9063074588775635 ] }, "params_m": 3.877984, "gift_final": { "per_dataset": { "m4_weekly": { "CRPS": 0.12512168968327364, "MASE": 13.579908332826367, "freq": "W-SUN" }, "m4_daily": { "CRPS": 0.10828313369755001, "MASE": 16.003181458704287, "freq": "D" }, "m4_hourly": { "CRPS": 0.11515012822351552, "MASE": 10.883395805916832, "freq": "H" }, "m4_monthly": { "CRPS": 0.1450549697006775, "MASE": 1.775619489902685, "freq": "M" }, "electricity/15T": { "CRPS": 0.24833223717009067, "MASE": 2.917213232022309, "freq": "15T" }, "electricity/H": { "CRPS": 0.23775977993520833, "MASE": 3.3337107771795096, "freq": "H" }, "solar/10T": { "CRPS": 0.9811011581473177, "MASE": 1.532356273347243, "freq": "10T" }, "solar/H": { "CRPS": 0.9937472405733553, "MASE": 2.105729410830325, "freq": "H" }, "ett1/15T": { "CRPS": 0.38953334817121527, "MASE": 1.5270910506730995, "freq": "15T" }, "ett2/H": { "CRPS": 0.1011386396260841, "MASE": 1.084675674620597, "freq": "H" }, "jena_weather/10T": { "CRPS": 0.0579320925156447, "MASE": 0.4825792630509371, "freq": "10T" }, "us_births/D": { "CRPS": 0.07490032071819595, "MASE": 1.554195402465568, "freq": "D" }, "covid_deaths": { "CRPS": 0.13747515073753727, "MASE": 69.12288498792664, "freq": "D" }, "hospital": { "CRPS": 0.07045661221848139, "MASE": 0.8775696794960403, "freq": "M" } }, "gm_crps": 0.17303462645133794, "gm_crps_long_season": 0.2497329377428901, "gm_crps_other": 0.10609036526016245 }, "gift_7500": { "per_dataset": { "m4_weekly": { "CRPS": 0.16017692881872042, "MASE": 16.50203318990757, "freq": "W-SUN" }, "m4_daily": { "CRPS": 0.11852580410988955, "MASE": 16.881461683050247, "freq": "D" }, "m4_hourly": { "CRPS": 0.11931458576270611, "MASE": 11.024283032787515, "freq": "H" }, "m4_monthly": { "CRPS": 0.15958782035673627, "MASE": 1.935500072619252, "freq": "M" }, "electricity/15T": { "CRPS": 0.2546298576803023, "MASE": 2.948334053055504, "freq": "15T" }, "electricity/H": { "CRPS": 0.27087805546225513, "MASE": 3.4240670663379826, "freq": "H" }, "solar/10T": { "CRPS": 1.004875608561122, "MASE": 1.5289481339753446, "freq": "10T" }, "solar/H": { "CRPS": 0.9713012020902768, "MASE": 2.0984343175051787, "freq": "H" }, "ett1/15T": { "CRPS": 0.4003773777633688, "MASE": 1.6204739503578354, "freq": "15T" }, "ett2/H": { "CRPS": 0.10689069993944299, "MASE": 1.121414455826455, "freq": "H" }, "jena_weather/10T": { "CRPS": 0.059192942631696756, "MASE": 0.5608103852580025, "freq": "10T" }, "us_births/D": { "CRPS": 0.07447188207995353, "MASE": 1.5145054902207784, "freq": "D" }, "covid_deaths": { "CRPS": 0.28883141050357974, "MASE": 92.55579268397473, "freq": "D" }, "hospital": { "CRPS": 0.07459479258516391, "MASE": 0.8895246680750647, "freq": "M" } }, "gm_crps": 0.19291995663264858, "gm_crps_long_season": 0.25915839859359296, "gm_crps_other": 0.13015463729013377 }, "gift_15000": { "per_dataset": { "m4_weekly": { "CRPS": 0.22020264148775712, "MASE": 23.698480104605164, "freq": "W-SUN" }, "m4_daily": { "CRPS": 0.11880221788693691, "MASE": 15.140293271077125, "freq": "D" }, "m4_hourly": { "CRPS": 0.12283461463180001, "MASE": 10.77291666051461, "freq": "H" }, "m4_monthly": { "CRPS": 0.21153679853307372, "MASE": 2.201986729166534, "freq": "M" }, "electricity/15T": { "CRPS": 0.24194931981236845, "MASE": 2.876919897927593, "freq": "15T" }, "electricity/H": { "CRPS": 0.2663944759695252, "MASE": 3.3960649194206005, "freq": "H" }, "solar/10T": { "CRPS": 0.9770156337969402, "MASE": 1.5398334393585549, "freq": "10T" }, "solar/H": { "CRPS": 1.0147348104783225, "MASE": 2.1719436959628644, "freq": "H" }, "ett1/15T": { "CRPS": 0.38676366554678254, "MASE": 1.5525926584928091, "freq": "15T" }, "ett2/H": { "CRPS": 0.11291669343691307, "MASE": 1.1611004643171088, "freq": "H" }, "jena_weather/10T": { "CRPS": 0.06008666365339678, "MASE": 0.574510818232375, "freq": "10T" }, "us_births/D": { "CRPS": 0.07541937141300581, "MASE": 1.49046327425222, "freq": "D" }, "covid_deaths": { "CRPS": 0.2030276417172036, "MASE": 90.59166758487763, "freq": "D" }, "hospital": { "CRPS": 0.0837334066866807, "MASE": 0.9020981896013266, "freq": "M" } }, "gm_crps": 0.198373664037254, "gm_crps_long_season": 0.25955284660327843, "gm_crps_other": 0.138620769892888 }, "gift_22500": { "per_dataset": { "m4_weekly": { "CRPS": 0.1265290884794003, "MASE": 13.073245795214634, "freq": "W-SUN" }, "m4_daily": { "CRPS": 0.1014370602161519, "MASE": 14.449099730840066, "freq": "D" }, "m4_hourly": { "CRPS": 0.12055590421871314, "MASE": 10.707814174080637, "freq": "H" }, "m4_monthly": { "CRPS": 0.1646702685525826, "MASE": 1.9805286059122045, "freq": "M" }, "electricity/15T": { "CRPS": 0.2537406635063478, "MASE": 2.9660510419923054, "freq": "15T" }, "electricity/H": { "CRPS": 0.2499510808111591, "MASE": 3.353257908311701, "freq": "H" }, "solar/10T": { "CRPS": 0.9987311647153546, "MASE": 1.5202613786851409, "freq": "10T" }, "solar/H": { "CRPS": 0.9966099131105831, "MASE": 2.0968730692794755, "freq": "H" }, "ett1/15T": { "CRPS": 0.39966744584625125, "MASE": 1.563963666295292, "freq": "15T" }, "ett2/H": { "CRPS": 0.1011082142881225, "MASE": 1.1142797604876018, "freq": "H" }, "jena_weather/10T": { "CRPS": 0.05807331923285683, "MASE": 0.5124247205781431, "freq": "10T" }, "us_births/D": { "CRPS": 0.07536039959582448, "MASE": 1.5191226085850418, "freq": "D" }, "covid_deaths": { "CRPS": 0.1901402083964537, "MASE": 95.18876202752733, "freq": "D" }, "hospital": { "CRPS": 0.09080732090363478, "MASE": 1.01370677756089, "freq": "M" } }, "gm_crps": 0.1835091911117041, "gm_crps_long_season": 0.2549668130353086, "gm_crps_other": 0.11836501970607557 }, "probe": { "pearson": [ 0.065, -0.011, -0.011, -0.001, -0.003, -0.014, -0.016, -0.006 ], "amp_retention": [ 0.194, 0.237, 0.217, 0.207, 0.208, 0.207, 0.208, 0.209 ] } }